xref: /openbmc/qemu/include/block/block.h (revision 88c725c7)
1 #ifndef BLOCK_H
2 #define BLOCK_H
3 
4 #include "block/aio.h"
5 #include "qemu/iov.h"
6 #include "qemu/option.h"
7 #include "qemu/coroutine.h"
8 #include "block/accounting.h"
9 #include "block/dirty-bitmap.h"
10 #include "block/blockjob.h"
11 #include "qapi/qmp/qobject.h"
12 #include "qapi-types.h"
13 #include "qemu/hbitmap.h"
14 
15 /* block.c */
16 typedef struct BlockDriver BlockDriver;
17 typedef struct BdrvChild BdrvChild;
18 typedef struct BdrvChildRole BdrvChildRole;
19 
20 typedef struct BlockDriverInfo {
21     /* in bytes, 0 if irrelevant */
22     int cluster_size;
23     /* offset at which the VM state can be saved (0 if not possible) */
24     int64_t vm_state_offset;
25     bool is_dirty;
26     /*
27      * True if unallocated blocks read back as zeroes. This is equivalent
28      * to the LBPRZ flag in the SCSI logical block provisioning page.
29      */
30     bool unallocated_blocks_are_zero;
31     /*
32      * True if the driver can optimize writing zeroes by unmapping
33      * sectors. This is equivalent to the BLKDISCARDZEROES ioctl in Linux
34      * with the difference that in qemu a discard is allowed to silently
35      * fail. Therefore we have to use bdrv_pwrite_zeroes with the
36      * BDRV_REQ_MAY_UNMAP flag for an optimized zero write with unmapping.
37      * After this call the driver has to guarantee that the contents read
38      * back as zero. It is additionally required that the block device is
39      * opened with BDRV_O_UNMAP flag for this to work.
40      */
41     bool can_write_zeroes_with_unmap;
42     /*
43      * True if this block driver only supports compressed writes
44      */
45     bool needs_compressed_writes;
46 } BlockDriverInfo;
47 
48 typedef struct BlockFragInfo {
49     uint64_t allocated_clusters;
50     uint64_t total_clusters;
51     uint64_t fragmented_clusters;
52     uint64_t compressed_clusters;
53 } BlockFragInfo;
54 
55 typedef enum {
56     BDRV_REQ_COPY_ON_READ       = 0x1,
57     BDRV_REQ_ZERO_WRITE         = 0x2,
58     /* The BDRV_REQ_MAY_UNMAP flag is used to indicate that the block driver
59      * is allowed to optimize a write zeroes request by unmapping (discarding)
60      * blocks if it is guaranteed that the result will read back as
61      * zeroes. The flag is only passed to the driver if the block device is
62      * opened with BDRV_O_UNMAP.
63      */
64     BDRV_REQ_MAY_UNMAP          = 0x4,
65     BDRV_REQ_NO_SERIALISING     = 0x8,
66     BDRV_REQ_FUA                = 0x10,
67     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
68 
69     /* Mask of valid flags */
70     BDRV_REQ_MASK               = 0x3f,
71 } BdrvRequestFlags;
72 
73 typedef struct BlockSizes {
74     uint32_t phys;
75     uint32_t log;
76 } BlockSizes;
77 
78 typedef struct HDGeometry {
79     uint32_t heads;
80     uint32_t sectors;
81     uint32_t cylinders;
82 } HDGeometry;
83 
84 #define BDRV_O_RDWR        0x0002
85 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
86 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
87 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
88 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
89 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
90 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
91 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
92 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
93 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
94 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
95 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
96 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
97 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
98                                       select an appropriate protocol driver,
99                                       ignoring the format layer */
100 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
101 
102 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
103 
104 
105 /* Option names of options parsed by the block layer */
106 
107 #define BDRV_OPT_CACHE_WB       "cache.writeback"
108 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
109 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
110 #define BDRV_OPT_READ_ONLY      "read-only"
111 #define BDRV_OPT_DISCARD        "discard"
112 #define BDRV_OPT_FORCE_SHARE    "force-share"
113 
114 
115 #define BDRV_SECTOR_BITS   9
116 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
117 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
118 
119 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
120                                      INT_MAX >> BDRV_SECTOR_BITS)
121 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
122 
123 /*
124  * Allocation status flags for bdrv_get_block_status() and friends.
125  *
126  * Public flags:
127  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
128  * BDRV_BLOCK_ZERO: offset reads as zero
129  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
130  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
131  *                       layer (short for DATA || ZERO), set by block layer
132  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this layer
133  *
134  * Internal flag:
135  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
136  *                 that the block layer recompute the answer from the returned
137  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
138  *
139  * If BDRV_BLOCK_OFFSET_VALID is set, bits 9-62 (BDRV_BLOCK_OFFSET_MASK)
140  * represent the offset in the returned BDS that is allocated for the
141  * corresponding raw data; however, whether that offset actually contains
142  * data also depends on BDRV_BLOCK_DATA and BDRV_BLOCK_ZERO, as follows:
143  *
144  * DATA ZERO OFFSET_VALID
145  *  t    t        t       sectors read as zero, returned file is zero at offset
146  *  t    f        t       sectors read as valid from file at offset
147  *  f    t        t       sectors preallocated, read as zero, returned file not
148  *                        necessarily zero at offset
149  *  f    f        t       sectors preallocated but read from backing_hd,
150  *                        returned file contains garbage at offset
151  *  t    t        f       sectors preallocated, read as zero, unknown offset
152  *  t    f        f       sectors read from unknown file or offset
153  *  f    t        f       not allocated or unknown offset, read as zero
154  *  f    f        f       not allocated or unknown offset, read from backing_hd
155  */
156 #define BDRV_BLOCK_DATA         0x01
157 #define BDRV_BLOCK_ZERO         0x02
158 #define BDRV_BLOCK_OFFSET_VALID 0x04
159 #define BDRV_BLOCK_RAW          0x08
160 #define BDRV_BLOCK_ALLOCATED    0x10
161 #define BDRV_BLOCK_EOF          0x20
162 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
163 
164 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
165 
166 typedef struct BDRVReopenState {
167     BlockDriverState *bs;
168     int flags;
169     QDict *options;
170     QDict *explicit_options;
171     void *opaque;
172 } BDRVReopenState;
173 
174 /*
175  * Block operation types
176  */
177 typedef enum BlockOpType {
178     BLOCK_OP_TYPE_BACKUP_SOURCE,
179     BLOCK_OP_TYPE_BACKUP_TARGET,
180     BLOCK_OP_TYPE_CHANGE,
181     BLOCK_OP_TYPE_COMMIT_SOURCE,
182     BLOCK_OP_TYPE_COMMIT_TARGET,
183     BLOCK_OP_TYPE_DATAPLANE,
184     BLOCK_OP_TYPE_DRIVE_DEL,
185     BLOCK_OP_TYPE_EJECT,
186     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
187     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
188     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
189     BLOCK_OP_TYPE_MIRROR_SOURCE,
190     BLOCK_OP_TYPE_MIRROR_TARGET,
191     BLOCK_OP_TYPE_RESIZE,
192     BLOCK_OP_TYPE_STREAM,
193     BLOCK_OP_TYPE_REPLACE,
194     BLOCK_OP_TYPE_MAX,
195 } BlockOpType;
196 
197 /* Block node permission constants */
198 enum {
199     /**
200      * A user that has the "permission" of consistent reads is guaranteed that
201      * their view of the contents of the block device is complete and
202      * self-consistent, representing the contents of a disk at a specific
203      * point.
204      *
205      * For most block devices (including their backing files) this is true, but
206      * the property cannot be maintained in a few situations like for
207      * intermediate nodes of a commit block job.
208      */
209     BLK_PERM_CONSISTENT_READ    = 0x01,
210 
211     /** This permission is required to change the visible disk contents. */
212     BLK_PERM_WRITE              = 0x02,
213 
214     /**
215      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
216      * required for writes to the block node when the caller promises that
217      * the visible disk content doesn't change.
218      */
219     BLK_PERM_WRITE_UNCHANGED    = 0x04,
220 
221     /** This permission is required to change the size of a block node. */
222     BLK_PERM_RESIZE             = 0x08,
223 
224     /**
225      * This permission is required to change the node that this BdrvChild
226      * points to.
227      */
228     BLK_PERM_GRAPH_MOD          = 0x10,
229 
230     BLK_PERM_ALL                = 0x1f,
231 };
232 
233 char *bdrv_perm_names(uint64_t perm);
234 
235 /* disk I/O throttling */
236 void bdrv_init(void);
237 void bdrv_init_with_whitelist(void);
238 bool bdrv_uses_whitelist(void);
239 BlockDriver *bdrv_find_protocol(const char *filename,
240                                 bool allow_protocol_prefix,
241                                 Error **errp);
242 BlockDriver *bdrv_find_format(const char *format_name);
243 int bdrv_create(BlockDriver *drv, const char* filename,
244                 QemuOpts *opts, Error **errp);
245 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
246 BlockDriverState *bdrv_new(void);
247 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
248                  Error **errp);
249 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
250                        Error **errp);
251 
252 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
253 int bdrv_parse_discard_flags(const char *mode, int *flags);
254 BdrvChild *bdrv_open_child(const char *filename,
255                            QDict *options, const char *bdref_key,
256                            BlockDriverState* parent,
257                            const BdrvChildRole *child_role,
258                            bool allow_none, Error **errp);
259 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
260                          Error **errp);
261 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
262                            const char *bdref_key, Error **errp);
263 BlockDriverState *bdrv_open(const char *filename, const char *reference,
264                             QDict *options, int flags, Error **errp);
265 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
266                                        int flags, Error **errp);
267 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
268                                     BlockDriverState *bs,
269                                     QDict *options, int flags);
270 int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp);
271 int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp);
272 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
273                         BlockReopenQueue *queue, Error **errp);
274 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
275 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
276 int bdrv_read(BdrvChild *child, int64_t sector_num,
277               uint8_t *buf, int nb_sectors);
278 int bdrv_write(BdrvChild *child, int64_t sector_num,
279                const uint8_t *buf, int nb_sectors);
280 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
281                        int bytes, BdrvRequestFlags flags);
282 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
283 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
284 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
285 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
286 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
287 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
288                      const void *buf, int count);
289 int coroutine_fn bdrv_co_readv(BdrvChild *child, int64_t sector_num,
290                                int nb_sectors, QEMUIOVector *qiov);
291 int coroutine_fn bdrv_co_writev(BdrvChild *child, int64_t sector_num,
292                                int nb_sectors, QEMUIOVector *qiov);
293 /*
294  * Efficiently zero a region of the disk image.  Note that this is a regular
295  * I/O request like read or write and should have a reasonable size.  This
296  * function is not suitable for zeroing the entire image in a single request
297  * because it may allocate memory for the entire region.
298  */
299 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
300                                        int bytes, BdrvRequestFlags flags);
301 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
302     const char *backing_file);
303 void bdrv_refresh_filename(BlockDriverState *bs);
304 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
305                   Error **errp);
306 int64_t bdrv_nb_sectors(BlockDriverState *bs);
307 int64_t bdrv_getlength(BlockDriverState *bs);
308 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
309 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
310                                BlockDriverState *in_bs, Error **errp);
311 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
312 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
313 int bdrv_commit(BlockDriverState *bs);
314 int bdrv_change_backing_file(BlockDriverState *bs,
315     const char *backing_file, const char *backing_fmt);
316 void bdrv_register(BlockDriver *bdrv);
317 int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top,
318                            BlockDriverState *base,
319                            const char *backing_file_str);
320 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
321                                     BlockDriverState *bs);
322 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
323 
324 
325 typedef struct BdrvCheckResult {
326     int corruptions;
327     int leaks;
328     int check_errors;
329     int corruptions_fixed;
330     int leaks_fixed;
331     int64_t image_end_offset;
332     BlockFragInfo bfi;
333 } BdrvCheckResult;
334 
335 typedef enum {
336     BDRV_FIX_LEAKS    = 1,
337     BDRV_FIX_ERRORS   = 2,
338 } BdrvCheckMode;
339 
340 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
341 
342 /* The units of offset and total_work_size may be chosen arbitrarily by the
343  * block driver; total_work_size may change during the course of the amendment
344  * operation */
345 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
346                                       int64_t total_work_size, void *opaque);
347 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
348                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque);
349 
350 /* external snapshots */
351 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
352                                       BlockDriverState *candidate);
353 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
354 
355 /* check if a named node can be replaced when doing drive-mirror */
356 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
357                                         const char *node_name, Error **errp);
358 
359 /* async block I/O */
360 void bdrv_aio_cancel(BlockAIOCB *acb);
361 void bdrv_aio_cancel_async(BlockAIOCB *acb);
362 
363 /* sg packet commands */
364 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
365 
366 /* Invalidate any cached metadata used by image formats */
367 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
368 void bdrv_invalidate_cache_all(Error **errp);
369 int bdrv_inactivate_all(void);
370 
371 /* Ensure contents are flushed to disk.  */
372 int bdrv_flush(BlockDriverState *bs);
373 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
374 int bdrv_flush_all(void);
375 void bdrv_close_all(void);
376 void bdrv_drain(BlockDriverState *bs);
377 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
378 void bdrv_drain_all_begin(void);
379 void bdrv_drain_all_end(void);
380 void bdrv_drain_all(void);
381 
382 #define BDRV_POLL_WHILE(bs, cond) ({                       \
383     bool waited_ = false;                                  \
384     bool busy_ = true;                                     \
385     BlockDriverState *bs_ = (bs);                          \
386     AioContext *ctx_ = bdrv_get_aio_context(bs_);          \
387     if (aio_context_in_iothread(ctx_)) {                   \
388         while ((cond) || busy_) {                          \
389             busy_ = aio_poll(ctx_, (cond));                \
390             waited_ |= !!(cond) | busy_;                   \
391         }                                                  \
392     } else {                                               \
393         assert(qemu_get_current_aio_context() ==           \
394                qemu_get_aio_context());                    \
395         /* Ask bdrv_dec_in_flight to wake up the main      \
396          * QEMU AioContext.  Extra I/O threads never take  \
397          * other I/O threads' AioContexts (see for example \
398          * block_job_defer_to_main_loop for how to do it). \
399          */                                                \
400         assert(!bs_->wakeup);                              \
401         /* Set bs->wakeup before evaluating cond.  */      \
402         atomic_mb_set(&bs_->wakeup, true);                 \
403         while (busy_) {                                    \
404             if ((cond)) {                                  \
405                 waited_ = busy_ = true;                    \
406                 aio_context_release(ctx_);                 \
407                 aio_poll(qemu_get_aio_context(), true);    \
408                 aio_context_acquire(ctx_);                 \
409             } else {                                       \
410                 busy_ = aio_poll(ctx_, false);             \
411                 waited_ |= busy_;                          \
412             }                                              \
413         }                                                  \
414         atomic_set(&bs_->wakeup, false);                   \
415     }                                                      \
416     waited_; })
417 
418 int bdrv_pdiscard(BlockDriverState *bs, int64_t offset, int bytes);
419 int bdrv_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes);
420 int bdrv_has_zero_init_1(BlockDriverState *bs);
421 int bdrv_has_zero_init(BlockDriverState *bs);
422 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
423 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
424 int64_t bdrv_get_block_status(BlockDriverState *bs, int64_t sector_num,
425                               int nb_sectors, int *pnum,
426                               BlockDriverState **file);
427 int64_t bdrv_get_block_status_above(BlockDriverState *bs,
428                                     BlockDriverState *base,
429                                     int64_t sector_num,
430                                     int nb_sectors, int *pnum,
431                                     BlockDriverState **file);
432 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
433                       int64_t *pnum);
434 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
435                             int64_t offset, int64_t bytes, int64_t *pnum);
436 
437 bool bdrv_is_read_only(BlockDriverState *bs);
438 bool bdrv_is_writable(BlockDriverState *bs);
439 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
440                            bool ignore_allow_rdw, Error **errp);
441 int bdrv_set_read_only(BlockDriverState *bs, bool read_only, Error **errp);
442 bool bdrv_is_sg(BlockDriverState *bs);
443 bool bdrv_is_inserted(BlockDriverState *bs);
444 int bdrv_media_changed(BlockDriverState *bs);
445 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
446 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
447 const char *bdrv_get_format_name(BlockDriverState *bs);
448 BlockDriverState *bdrv_find_node(const char *node_name);
449 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
450 BlockDriverState *bdrv_lookup_bs(const char *device,
451                                  const char *node_name,
452                                  Error **errp);
453 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
454 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
455 
456 typedef struct BdrvNextIterator {
457     enum {
458         BDRV_NEXT_BACKEND_ROOTS,
459         BDRV_NEXT_MONITOR_OWNED,
460     } phase;
461     BlockBackend *blk;
462     BlockDriverState *bs;
463 } BdrvNextIterator;
464 
465 BlockDriverState *bdrv_first(BdrvNextIterator *it);
466 BlockDriverState *bdrv_next(BdrvNextIterator *it);
467 
468 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
469 bool bdrv_is_encrypted(BlockDriverState *bs);
470 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
471                          void *opaque);
472 const char *bdrv_get_node_name(const BlockDriverState *bs);
473 const char *bdrv_get_device_name(const BlockDriverState *bs);
474 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
475 int bdrv_get_flags(BlockDriverState *bs);
476 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
477 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs);
478 void bdrv_round_to_clusters(BlockDriverState *bs,
479                             int64_t offset, unsigned int bytes,
480                             int64_t *cluster_offset,
481                             unsigned int *cluster_bytes);
482 
483 const char *bdrv_get_encrypted_filename(BlockDriverState *bs);
484 void bdrv_get_backing_filename(BlockDriverState *bs,
485                                char *filename, int filename_size);
486 void bdrv_get_full_backing_filename(BlockDriverState *bs,
487                                     char *dest, size_t sz, Error **errp);
488 void bdrv_get_full_backing_filename_from_filename(const char *backed,
489                                                   const char *backing,
490                                                   char *dest, size_t sz,
491                                                   Error **errp);
492 
493 int path_has_protocol(const char *path);
494 int path_is_absolute(const char *path);
495 void path_combine(char *dest, int dest_size,
496                   const char *base_path,
497                   const char *filename);
498 
499 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
500 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
501 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
502                       int64_t pos, int size);
503 
504 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
505                       int64_t pos, int size);
506 
507 void bdrv_img_create(const char *filename, const char *fmt,
508                      const char *base_filename, const char *base_fmt,
509                      char *options, uint64_t img_size, int flags,
510                      bool quiet, Error **errp);
511 
512 /* Returns the alignment in bytes that is required so that no bounce buffer
513  * is required throughout the stack */
514 size_t bdrv_min_mem_align(BlockDriverState *bs);
515 /* Returns optimal alignment in bytes for bounce buffer */
516 size_t bdrv_opt_mem_align(BlockDriverState *bs);
517 void *qemu_blockalign(BlockDriverState *bs, size_t size);
518 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
519 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
520 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
521 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
522 
523 void bdrv_enable_copy_on_read(BlockDriverState *bs);
524 void bdrv_disable_copy_on_read(BlockDriverState *bs);
525 
526 void bdrv_ref(BlockDriverState *bs);
527 void bdrv_unref(BlockDriverState *bs);
528 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
529 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
530                              BlockDriverState *child_bs,
531                              const char *child_name,
532                              const BdrvChildRole *child_role,
533                              Error **errp);
534 
535 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
536 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
537 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
538 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
539 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
540 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
541 
542 #define BLKDBG_EVENT(child, evt) \
543     do { \
544         if (child) { \
545             bdrv_debug_event(child->bs, evt); \
546         } \
547     } while (0)
548 
549 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
550 
551 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
552                            const char *tag);
553 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
554 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
555 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
556 
557 /**
558  * bdrv_get_aio_context:
559  *
560  * Returns: the currently bound #AioContext
561  */
562 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
563 
564 /**
565  * Transfer control to @co in the aio context of @bs
566  */
567 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
568 
569 /**
570  * bdrv_set_aio_context:
571  *
572  * Changes the #AioContext used for fd handlers, timers, and BHs by this
573  * BlockDriverState and all its children.
574  *
575  * This function must be called with iothread lock held.
576  */
577 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
578 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
579 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
580 
581 void bdrv_io_plug(BlockDriverState *bs);
582 void bdrv_io_unplug(BlockDriverState *bs);
583 
584 /**
585  * bdrv_parent_drained_begin:
586  *
587  * Begin a quiesced section of all users of @bs. This is part of
588  * bdrv_drained_begin.
589  */
590 void bdrv_parent_drained_begin(BlockDriverState *bs);
591 
592 /**
593  * bdrv_parent_drained_end:
594  *
595  * End a quiesced section of all users of @bs. This is part of
596  * bdrv_drained_end.
597  */
598 void bdrv_parent_drained_end(BlockDriverState *bs);
599 
600 /**
601  * bdrv_drained_begin:
602  *
603  * Begin a quiesced section for exclusive access to the BDS, by disabling
604  * external request sources including NBD server and device model. Note that
605  * this doesn't block timers or coroutines from submitting more requests, which
606  * means block_job_pause is still necessary.
607  *
608  * This function can be recursive.
609  */
610 void bdrv_drained_begin(BlockDriverState *bs);
611 
612 /**
613  * bdrv_drained_end:
614  *
615  * End a quiescent section started by bdrv_drained_begin().
616  */
617 void bdrv_drained_end(BlockDriverState *bs);
618 
619 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
620                     Error **errp);
621 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
622 
623 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
624                                      uint32_t granularity, Error **errp);
625 
626 #endif
627